Blind Signal Separation in the Presence of Gaussian Noise
description
Transcript of Blind Signal Separation in the Presence of Gaussian Noise
PowerPoint Presentation
Mikhail BelkinLuis RademacherJames VossOhio State UniversityBlind Signal Separation in the Presence of Gaussian NoiseCocktail Party Problem (Example)
Independent Component Analysis (ICA)
Typical (noiseless) ICA ProcedureRaw DataWhitenRecover RotationDemixed DataData mixed from Uniform Distributions
Relaxing Step 1Whitened DataQuasi-Whitened DataRelated Work (efficient noisy ICA)Aapo Hyvrinen (1999) discusses noisy ICA when the noise covariance is known.Arie Yeredor (2000) provides a one-step solution to noisy ICA using the Hessian of the directional 2nd Characteristic Function.Arora, Ge, Moitra, and Sachdeva (2012) introduced quasi-whitening and provide an efficient noisy ICA algorithm for the special case where all latent signals have fourth cumulant of the same sign.Hsu and Kakade (2012) state a one-step solution to noisy ICA using the Hessian of the directional fourth cumulant.Our ContributionWhat are Cumulants?Properties of Multivariate Cumulants(stated for fourth cumulant)Quasi-Whitening Algorithm
Algorithms ValidityAlgorithms ValidityMain ResultQuasi-Whitening Algorithm Restated
Provably efficientPerforms the relaxed Step 1 of noisy ICA (quasi-whitening).Compatible with small variations on existing algorithms for Step 2 of ICA.Thank YouAny Questions?